DRG入组付费异常预警模型研究的系统评价综述  

A Systematic Review of Research on DRG Payment Abnormal Warning Model

作　　者：刘硕 杨志平 伍俊 李顺飞[3] 丁敬美 赵莉 胡雪军 

机构地区：[1]中国人民解放军空军军医大学第一附属医院,西安710032 [2]中国人民解放军空军军医大学,西安710032 [3]中国人民解放军总医院,北京100036 [4]中国人民解放军中部战区总医院,武汉430012

出　　处：《中国医疗保险》2025年第3期60-67,共8页China Health Insurance

基　　金：国家自然科学基金项目“基于深度学习挖掘电子病历非结构化数据的医保DRGs病例入组付费异常的智能预警模型及策略研究”(72474222)。

摘　　要：目的:为解决DRG低码高编、高码低编、歧义病例等入组付费异常问题,本研究对国内外DRG入组付费异常相关模型文章开展系统性评价综述。方法:本文运用PRISMA综述框架对文献报告质量进行评估,对文献模型的研究设计、总体趋势、数据来源、输入输出变量、评价指标等方面进行描述性分析。结果:总结了六类文献的研究设计类型,发现深度学习方法自2020年起在国外迅速发展,具有多中心数据来源、文本型输入变量、分类输出变量、模型评价性能高等显著特征,并提炼了科学问题、模型特点和局限性。结论:基于电子病历文本数据的全病组、高精度的DRG入组付费异常预警模型的深度学习研究较为缺乏,未来仍需不断探索基于深度学习电子病历结构化与非结构化数据的DRG入组付费异常预警工具,以进一步推动DRG支付方式改革向纵深发展。Objective:In order to solve the problems of DRG abnormal grouping such as upcoding,downcoding and ambiguous cases,a systematic review was carried out on the relevant articles on DRG abnormal grouping models at home and abroad.Methods:The PRISMA framework was used to evaluate the quality of literature and reports,and the research design,overall trends,data sources,input and output variables,evaluation indicators and other aspects of literature models were analyzed.Results:The research design types of the six categories of literature were summarized,and it was found that deep learning methods had developed rapidly in foreign countries since 2020,with significant features such as multi-center data sources,text-based input variables,output variables by type,and high model evaluation performance.Besides,scientific problems,model characteristics and limitations were extracted.Conclusion:There is a lack of research on whole-disease group and high-precision DRG payment abnormal warning models based on electronic medical record text data with deep learning methods.In the future,continuous exploration is needed to develop DRG payment abnormal warning tools based on deep learning of electronic medical record structured and unstructured data,in order to further promote the deepening of DRG payment reform.

关 键 词：DRG 异常入组 预警模型 系统评价 

分 类 号：F840.684[经济管理—保险] C913.7[社会学]